Annonce

annonce@lrde.epita.fr

564 discussions

by Alexandre Duret-Lutz

We are happy to announce the release of Spot 2.2. You can find the new release here: http://www.lrde.epita.fr/dload/spot/spot-2.2.tar.gz See https://spot.lrde.epita.fr/ for documentation and installation instructions. Spot 2.2 fixes some serious bugs and adds an assortments of small improvements. These changes have been contributed by Alexandre Gbaguidi Aisse, Étienne Renault, and myself. A detailed list of what is new is given below. As always, please direct any feedback to <spot(a)lrde.epita.fr>. New in spot 2.2 (2016-11-14) Command-line tools: * ltlfilt has a new option --from-ltlf to help reducing LTLf (i.e., LTL over finite words) model checking to LTL model checking. This is based on a transformation by De Giacomo & Vardi (IJCAI'13). * "ltldo --stats=%R", which used to display the serial number of the formula processed, was renamed to "ltldo --stats=%#" to free %R for the following feature. * autfilt, dstar2tgba, ltl2tgba, ltlcross, ltldo learned to measure cpu-time (as opposed to wall-time) using --stats=%R. User or system time, for children or parent, can be measured separately by adding additional %[LETTER]R options. The difference between %r (wall-clock time) and %R (CPU time) can also be used to detect unreliable measurements. See https://spot.lrde.epita.fr/oaut.html#timing Library: * from_ltlf() is a new function implementing the --from-ltlf transformation described above. * is_unambiguous() was rewritten in a more efficient way. * scc_info learned to determine the acceptance of simple SCCs made of a single self-loop without resorting to remove_fin() for complex acceptance conditions. * remove_fin() has been improved to better deal with automata with "unclean" acceptance, i.e., acceptance sets that are declared but not used. In particular, this helps scc_info to be more efficient at deciding the acceptance of SCCs in presence of Fin acceptance. * Simulation-based reductions now implement just bisimulation-based reductions on deterministic automata, to save time. As an example, this halves the run time of genltl --rv-counter=10 | ltl2tgba * scc_filter() learned to preserve state names and highlighted states. * The BuDDy library has been slightly optimized: the initial setup of the unicity table can now be vectorized by GCC, and all calls to setjmp are now avoided when variable reordering is disabled (the default). * The twa class has three new methods: aut->intersects(other) aut->intersecting_run(other) aut->intersecting_word(other) currently these are just convenient wrappers around !otf_product(aut, other)->is_empty() otf_product(aut, other)->accepting_run()->project(aut) otf_product(aut, other)->accepting_word() with any Fin-acceptance removal performed before the product. However the plan is to implement these more efficiently in the future. Bugs: * ltl2tgba was always using the highest settings for the LTL simplifier, ignoring the --low and --medium options. Now genltl --go-theta=12 | ltl2tgba --low --any is instantaneous as it should be. * The int-vector compression code could encode 6 and 22 in the same way, causing inevitable issues in our LTSmin interface. (This affects tests/ltsmin/modelcheck with option -z, not -Z.) * str_sere() and str_utf8_sere() were not returning the same string that print_sere() and print_utf8_sere() would print. * Running the LTL parser in debug mode would crash. * tgba_determinize() could produce incorrect deterministic automata when run with use_simulation=true (the default) on non-simplified automata. * When the automaton_stream_parser reads an automaton from an already opened file descriptor, it will not close the file anymore. Before that the spot.automata() python function used to close a file descriptor twice when reading from a pipe, and this led to crashes of the 0MQ library used by Jupyter, killing the Python kernel. * remove_fin() could produce incorrect result on incomplete automata tagged as weak and deterministic. * calling set_acceptance() several times on an automaton could result in unexpected behaviors, because set_acceptance(0,...) used to set the state-based acceptance flag automatically. * Some buffering issue caused syntax errors to be displayed out of place by the on-line translator. -- Alexandre Duret-Lutz

8 years, 4 months

10th European Lisp Symposium, April 3-4 2017, Brussels, Belgium

by Didier Verna

ELS'17 - 10th European Lisp Symposium VUB - Vrije Universiteit Brussel Belgium April 3-4, 2017 In co-location with <Programming> 2017 http://www.european-lisp-symposium.org/ The purpose of the European Lisp Symposium is to provide a forum for the discussion and dissemination of all aspects of design, implementation and application of any of the Lisp and Lisp-inspired dialects, including Common Lisp, Scheme, Emacs Lisp, AutoLisp, ISLISP, Dylan, Clojure, ACL2, ECMAScript, Racket, SKILL, Hop and so on. We encourage everyone interested in Lisp to participate. The 10th European Lisp Symposium invites high quality papers about novel research results, insights and lessons learned from practical applications and educational perspectives. We also encourage submissions about known ideas as long as they are presented in a new setting and/or in a highly elegant way. Topics include but are not limited to: - Context-, aspect-, domain-oriented and generative programming - Macro-, reflective-, meta- and/or rule-based development approaches - Language design and implementation - Language integration, inter-operation and deployment - Development methodologies, support and environments - Educational approaches and perspectives - Experience reports and case studies We invite submissions in the following forms: Papers: Technical papers of up to 8 pages that describe original results or explain known ideas in new and elegant ways. Demonstrations: Abstracts of up to 2 pages for demonstrations of tools, libraries, and applications. Tutorials: Abstracts of up to 4 pages for in-depth presentations about topics of special interest for at least 90 minutes and up to 180 minutes. The symposium will also provide slots for lightning talks, to be registered on-site every day. All submissions should be formatted following the ACM SIGS guidelines and include ACM classification categories and terms. For more information on the submission guidelines and the ACM keywords, see: http://www.acm.org/sigs/publications/proceedings-templates and http://www.acm.org/about/class/1998. The conference proceedings will be published in the ACM Digital Library. Important dates: - 30 Jan 2017 Submission deadline - 27 Feb 2017 Notification of acceptance - 20 Mar 2017 Final papers due - 03-04 Apr 2017 Symposium Programme chair: Alberto Riva, University of Florida, USA Programme committee: tba Search Keywords: #els2017, ELS 2017, ELS '17, European Lisp Symposium 2017, European Lisp Symposium '17, 10th ELS, 10th European Lisp Symposium, European Lisp Conference 2017, European Lisp Conference '17 -- Resistance is futile. You will be jazzimilated. Lisp, Jazz, Aïkido: http://www.didierverna.info

8 years, 4 months

Séminaire au LRDE le Mercredi 28 septembre 2016 : Samuel Charron

by Daniela Becker

Chers collègues, La prochaine session du séminaire du LRDE (Laboratoire de Recherche et Développement de l'EPITA) aura lieu le Mercredi 28 septembre 2016 (11h--12h), Salle L0 du LRDE. Vous trouverez sur le site du séminaire [1] les prochaines séances, les résumés, captations vidéos et planches des exposés précédents [2], le détail de cette séance [3] ainsi que le plan d'accès [4]. [1] http://seminaire.lrde.epita.fr [2] http://seminaire.lrde.epita.fr/Archives [3] http://seminaire.lrde.epita.fr/2016-09-28 [4] http://www.lrde.epita.fr/Contact Au programme du Mercredi 28 septembre 2016 : * 11h--12h: Transformation de la prospection commerciale grâce à la science des données -- Samuel Charron http://samuelcharron.fr Traditionnellement, la prospection commerciale B2B se fait grâce à des listes d'entreprises classées manuellement, sur la base de données administratives renseignées à la création des entreprises. Ces listes sont donc souvent dépassées, très chères et peu qualifiées. C-Radar, produit développé par la société Data Publica, veut transformer la prospection commerciale en enrichissant ces données administratives par des données issues du web. Ainsi, on obtient des données fraîches, plus ciblées. Et grâce aux techniques de machine learning et de clustering, on peut obtenir automatiquement des listes d'entreprises pertinentes à faible coût. Lors de cette présentation, nous aborderons les différentes techniques de science des données mises en œuvre dans ce produit, en relation avec les fonctionnalités du produit. -- Diplomé CSI 2008, Samuel Charron a travaillé chez différents éditeurs logiciels, toujours autour du thème de l'exploitation et de la valorisation des données issues du web. L'entrée du séminaire est libre. Merci de bien vouloir diffuser cette information le plus largement possible. N'hésitez pas à nous faire parvenir vos suggestions d’orateurs. -- Akim Demaille Akim.Demaille(a)lrde.epita.fr

8 years, 6 months

Spot 2.1 was released

by Alexandre Duret-Lutz

We are happy to announce the release of Spot 2.1. Spot 2.1 is mainly a collection of small improvements that have accumulated since the release of Spot 2.0 four months ago. A detailed list of those is given below. You can find the new release here: http://www.lrde.epita.fr/dload/spot/spot-2.1.tar.gz See https://spot.lrde.epita.fr/ for documentation and installation instructions. As always, please direct any feedback to <spot(a)lrde.epita.fr>. New in spot 2.1 (2016-08-08) Command-line tools: * All tools that input formulas or automata (i.e., autfilt, dstar2tgba, ltl2tgba, ltl2tgta, ltlcross, ltldo, ltlfilt, ltlgrind) now have a more homogeneous handling of the default input. - If no formula/automaton have been specified, and the standard input is a not a tty, then the default is to read that. This is a change for ltl2tgba and ltl2tgta. In particular, it simplifies genltl --dac | ltl2tgba -F- | autfilt ... into genltl --dac | ltl2tgba | autfilt ... - If standard input is a tty and no other input has been specified, then an error message is printed. This is a change for autfilt, dstar2tgba, ltlcross, ltldo, ltlfilt, ltlgrind, that used to expect the user to type formula or automata at the terminal, confusing people. - All tools now accept - as a shorthand for -F-, to force reading input from the standard input (regardless of whether it is a tty or not). This is a change for ltl2tgba, ltl2tgta, ltlcross, and ltldo. * ltldo has a new option --errors=... to specify how to deal with errors from executed tools. * ltlcross and ltldo learned to bypass the shell when executing simple commands (with support for single or double-quoted arguments, and redirection of stdin and stdout, but nothing more). * ltlcross and ltldo learned a new syntax to specify that an input formula should be written in some given syntax after rewriting some operators away. For instance the defaults arguments passed to ltl2dstar have been changed from --output-format=hoa %L %O into --output-format=hoa %[WM]L %O where [WM] specifies that operators W and M should be rewritten away. As a consequence running ltldo ltl2dstar -f 'a M b' will now work and call ltl2dstar on the equivalent formula 'b U (a & b)' instead. The operators that can be listed between brackets are the same as those of ltlfilt --unabbreviate option. * ltlcross learned to show some counterexamples when diagnosing failures of cross-comparison checks against random state spaces. * autfilt learned to filter automata by count of SCCs (--sccs=RANGE) or by type of SCCs (--accepting-sccs=RANGE, --rejecting-sccs=RANGE, trivial-sccs=RANGE, --terminal-sccs=RANGE, --weak-sccs=RANGE, --inherently-weak-sccs=RANGE). * autfilt learned --remove-unused-ap to remove atomic propositions that are declared in the input automaton, but not actually used. This of course makes sense only for input/output formats that declare atomic propositions (HOA & DSTAR). * autfilt learned two options to filter automata by count of used or unused atomic propositions: --used-ap=RANGE --unused-ap=RANGE. These differ from --ap=RANGE that only consider *declared* atomic propositions, regardless of whether they are actually used. * autfilt learned to filter automata by count of nondeterministsic states with --nondet-states=RANGE. * autfilt learned to filter automata representing stutter-invariant properties with --is-stutter-invariant. * autfilt learned two new options to highlight non-determinism: --highlight-nondet-states=NUM and --highlight-nondet-states=NUM where NUM is a color number. Additionally --highlight-nondet=NUM is a shorthand for using the two. * autfilt learned to highlight a run matching a given word using the --highlight-word=[NUM,]WORD option. However currently this only work on automata with Fin-less acceptance. * autfilt learned two options --generalized-rabin and --generalized-streett to convert the acceptance conditions. * genltl learned three new families: --dac-patterns=1..45, --eh-patterns=1..12, and --sb-patterns=1..27. Unlike other options these do not output scalable patterns, but simply a list of formulas appearing in these three papers: Dwyer et al (FMSP'98), Etessami & Holzmann (Concur'00), Somenzi & Bloem (CAV'00). * genltl learned two options, --positive and --negative, to control wether formulas should be output after negation or not (or both). * The formater used by --format (for ltlfilt, ltlgrind, genltl, randltl) or --stats (for autfilt, dstar2tgba, ltl2tgba, ltldo, randaut) learned to recognize double-quoted fields and double the double-quotes output inbetween as expected from RFC4180-compliant CSV files. For instance ltl2tgba -f 'a U "b+c"' --stats='"%f",%s' will output "a U ""b+c""",2 * The --csv-escape option of genltl, ltlfilt, ltlgrind, and randltl is now deprecated. The option is still here, but hidden and undocumented. * The --stats option of autfilt, dstar2tgba, ltl2tgba, ltldo, randaut learned to display the output (or input if that makes sense) automaton as a HOA one-liner using %h (or %H), helping to create CSV files contained automata. * autfilt and dstar2tgba learned to read automata from columns in CSV files, specified using the same filename/COLUMN syntax used by tools reading formulas. * Arguments passed to -x (in ltl2tgba, ltl2tgta, autfilt, dstar2tgba) that are not used are now reported as they might be typos. This ocurred a couple of times in our test-suite. A similar check is done for the arguments of autfilt --sat-minimize=... Library: * The print_hoa() function will now output version 1.1 of the HOA format when passed the "1.1" option (i.e., use -H1.1 from any command-line tool). As far as Spot is concerned, this allows negated properties to be expressed. Version 1 of the HOA format is still the default, but we plan to default to version 1.1 in the future. * The "highlight-states" and "highlight-edges" named properties, which were introduced in 1.99.8, will now be output using "spot.highlight.edges:" and "spot.highlight.states:" headers if version 1.1 of the HOA format is selected. The automaton parser was secretly able of reading that since 1.99.8, but that is now documented at https://spot.lrde.epita.fr/hoa.html#extensions * highlight_nondet_states() and highlight_nondet_edges() are new functions that define the above two named properties. * is_deterministic(), is_terminal(), is_weak(), and is_inherently_weak(), count_nondet_states(), highlight_nondet_edges(), highlight_nondet_states() will update the corresponding properties of the automaton as a side-effect of their check. * the sbacc() function, used by "ltl2tgba -S" and "autfilt -S" to convert automata to state-based acceptance, learned some tricks (using SCCs, pulling accepting marks common to all outgoing edges, and pushing acceptance marks common to all incoming edges) to reduce the number of additional states needed. * to_generalized_rabin() and to_generalized_streett() are two new functions that convert the acceptance condition as requested without changing the transition structure. * language_containment_checker now has default values for all parameters of its constructor. * spot::twa_run has a new method, project(), that can be used to project a run found in a product onto one of the original operand. This is for instance used by autfilt --highlight-word. The old spot::project_twa_run() function has been removed: it was not used anywhere in Spot, and had an obvious bug in its implementation, so it cannot be missed by anyone. * spot::twa has two new methods that supplement is_empty(): twa::accepting_run() and twa::accepting_word(). They compute what their names suggest. Note that twa::accepting_run(), unlike the two others, is currently restricted to automata with Fin-less acceptance. * spot::check_stutter_invariance() can now work on non-deterministic automata for which no corresponding formula is known. This implies that "autfilt --check=stutter" will now label all automata, not just deterministic automata. * New LTL and PSL simplification rules: - if e is pure eventuality and g => e, then e U g = Fg - if u is purely universal and u => g, then u R g = Gg - {s[*0..j]}[]->b = {s[*1..j]}[]->b - {s[*0..j]}<>->b = {s[*1..j]}<>->b * spot::twa::succ_iterable renamed to spot::internal::twa_succ_iterable to make it clear this is not for public consumption. * spot::fair_kripke::state_acceptance_conditions() renamed to spot::fair_kripke::state_acceptance_mark() for consistency. This is backward incompatible, but we are not aware of any actual use of this method. Python: * The __format__() method for formula supports the same operator-rewritting feature introduced in ltldo and ltlcross. So "{:[i]s}".format(f) is the same as "{:s}".format(f.unabbreviate("i")). * Bindings for language_containment_checker were added. * Bindings for randomize() were added. * Iterating over edges via "aut.out(s)" or "aut.edges()" now allows modifying the edge fields. * Under IPython the spot.ltsmin module now offers a %%pml magic to define promela models, compile them with spins, and dynamically load them. This is akin to the %%dve magic that was already supported. * The %%dve and %%pml magics honor the SPOT_TMPDIR and TMPDIR environment variables. This especially helps when the current directory is read-only. Documentation: * A new example page shows how to test the equivalence of two LTL/PSL formulas. https://spot.lrde.epita.fr/tut04.html * A new page discusses explicit vs. on-the-fly interfaces for exploring automata in C++. https://spot.lrde.epita.fr/tut50.html * Another new page shows how to implement an on-the-fly Kripke structure for a custom state space. https://spot.lrde.epita.fr/tut51.html * The concepts.html page now lists all named properties used by automata. Bug fixes: * When ltlcross found a bug using a product of complemented automata, the error message would report "Comp(Ni)*Comp(Pj)" as non-empty while the actual culprit was "Comp(Nj)*Comp(Pi)". * Fix some non-deterministic execution of minimize_wdba(), causing test-suite failures with the future G++ 7, and clang 3.9. * print_lbtt() had a memory leak when printing states without successors. -- Alexandre Duret-Lutz

8 years, 7 months

[Lrde] Publication accepted at IWRR-ECCV 2016

by Ana Stefania Calarasanu

I am happy to announce that the following paper has been accepted to the 2nd International Workshop on Robust Reading (IWRR), in conjunction with ECCV, that will take place in Amsterdam, on the 9th of September 2016: From text detection to text segmentation: a unified evaluation scheme Stefania Calarasanu (1), Jonathan Fabrizio (1) and Séverine Dubuisson (2) (1) LRDE-EPITA, 14-16, rue Voltaire, F-94276, Le Kremlin Bicêtre, France (2) CNRS, UMR 7222, ISIR, F-75005, Paris, France Abstract: Current text segmentation evaluation protocols are often incapable of properly handling different scenarios (broken/merged/partial characters). This leads to scores that incorrectly reflect the segmentation accuracy. In this article we propose a new evaluation scheme that overcomes most of the existent drawbacks by extending the EvaLTex protocol (initially designed to evaluate text detection at region level). This new unified platform has numerous advantages: it is able to evaluate a text understanding system at every detection stage and granularity level (paragraph/line/word and now character) by using the same metrics and matching rules; it is robust to all segmentation scenarios; it provides a qualitative and quantitative evaluation and a visual score representation that captures the whole behavior of a segmentation algorithm. Experimental results on nine segmentation algorithms using different evaluation frameworks are also provided to emphasize the interest of our method. Ana Stefania Calarasanu PhD Engineer - EPITA Research and Development Laboratory (LRDE) 14-16 rue Voltaire, 94276 Le Kremlin-Bicêtre CEDEX, France https://www.lrde.epita.fr/wiki/User:Calarasanu <https://www.lrde.epita.fr/wiki/User:Calarasanu>

8 years, 8 months

Vcsn 2.3 is released!

by Akim Demaille

About four hundred commits and five months after Vcsn 2.2, we are proud to announce the release of Vcsn 2.3, code-named "the tuple release"! http://vcsn.lrde.epita.fr/Vcsn2.3 As usual, many bugs were fixed (some quite old yet unnoticed so far!). Noteworthy changes include: - a particular effort was put on the documentation: there are thirty-five new documentation notebooks, and about forty others were improved. - full support for a "tuple" operator on all entities: expressions, polynomials, automata, etc. In [13]: aut = lambda e: vcsn.context('lan, q').expression(e).automaton() In [14]: a = aut('[ab]*') | aut('x') In [15]: a.shortest(6) Out[15]: \e|x + a|x + b|x + aa|x + ab|x + ba|x It is also available in the rational expressions themselves: In [16]: c = vcsn.context('lat<lan, lan>, q'); c Out[16]: {...}? x {...}? -> Q In [17]: e = c.expression('[ab]*|x'); e Out[17]: (a+b)*|x In [18]: e.shortest(6) Out[18]: \e|x + a|x + b|x + aa|x + ab|x + ba|x The derived-term algorithm supports this operator, and generates equivalent multitape automata. - many error messages were improved, to help users understand their mistakes. For instance, instead of In [2]: vcsn.Q.expression('a**').derivation('a') RuntimeError: q: star: invalid value: 1 we now display: In [2]: vcsn.Q.expression('a**').derivation('a') RuntimeError: Q: value is not starrable: 1 while computing derivative of: a** with respect to: a Besides, longuish stack traces under Python are now filtered from internal details. - in addition to `%automaton a`, which allows interactive edition of automata, the notebooks now feature two new interactive editors: `%context c` to edit/create context `c`, and `%expression e` for expressions (with an interactive display of the generated automata). - one may now generate random rational expressions and control the operators and their probabilities. - a lot of code improvement and consistency enforcement, both in C++ and in Python. For more details, please, see the news page (see below). People who worked on this release: - Akim Demaille - Clément Gillard - Lucien Boillod - Raoul Billion - Sébastien Piat - Thibaud Michaud People who influenced this release: - Alexandre Duret-Lutz - Jacques Sakarovitch - Luca Saiu - Sylvain Lombardy Vcsn 2.3: http://vcsn.lrde.epita.fr/Vcsn2.3 News page: http://vcsn.lrde.epita.fr/News_File Vcsn: http://vcsn.lrde.epita.fr Playground: http://vcsn-sandbox.lrde.epita.fr Doc: https://vcsn.lrde.epita.fr/dload/2.3/notebooks/!Read-me-first.html

8 years, 8 months

Le Rapport d'activité quinquennal du LRDE disponible

by Daniela Becker

Bonjour, Nous avons le plaisir de vous annoncer que le Rapport d’activité quinquennal du LRDE est désormais disponible sur notre site. Il retrace les activités du LRDE entre 2011 et 2015. Vous pouvez le télécharger ici : http://www.lrde.epita.fr/dload/rapport_activite/ra2015.pdf -- Daniela Becker Responsable administrative du LRDE

8 years, 8 months

New publication: ICTAC 2016

by Akim Demaille

We are pleased to announce that the following paper was accepted to the 13th International Colloquium on Theoretical Aspects of Computing (ICTAC 2016). Derived-term Automata for Extended Weighted Rational Expressions Akim Demaille EPITA/LRDE http://publis.lrde.epita.fr/demaille.16.ictac We present an algorithm to build an automaton from a rational expression. This approach introduces support for extended weighted expressions. Inspired by derived-term based algorithms, its core relies on a different construct, rational expansions. We introduce an inductive algorithm to compute the expansion of an expression from which the automaton follows. This algorithm is independent of the size of the alphabet, and actually even supports infinite alphabets. It can easily be accommodated to generate deterministic (weighted) automata. These constructs are implemented in Vcsn, a free-software platform dedicated to weighted automata and rational expressions.

8 years, 9 months

L'air de rien 36.0

by Daniela Becker

Bonjour, nous avons le plaisir de vous annoncer la sortie du n°36 du bulletin du LRDE. C'est un numéro spécial dédié aux séminaires des étudiants-chercheurs du LRDE qui auront lieu ce vendredi 1er juillet 2016. Vous y trouverez le programme de la journée avec les résumés des exposés. Nous y présentons également deux nouveaux projets auxquels participent les membres de l’équipe Image du LRDE. Vous pouvez télécharger le bulletin en couleur à la page suivante : http://www.lrde.epita.fr/wiki/LrdeBulletin/l-air-de-rien-36 -- Daniela Becker

8 years, 9 months

[Séminaire des étudiants CSI] Vendredi 1er juillet 2016 - Amphi Masters

by Daniela Becker

Bonjour, nous avons le plaisir de vous inviter au Séminaire des étudiants du LRDE. Il aura lieu le vendredi 1er juillet 2016 à partir de 10h00 en Amphi Masters (KB). -------------------------------------------- Au programme : VÉRIFICATION DU LOCUTEUR 10h00 : Compensation du Domain Mismatch pour la Reconnaissance du Locuteur – VALENTIN IOVENE Le domain mismatch se produit quand les données d’entraînement du système et les données utilisées pour l’évalution proviennent de sources différentes. Cette étude présente les techniques de pointe utilisées pour compenser le domain mismatch, comme une bibliothèque de blanchisseurs et l’utilisation d’une normalisation de la covariance invariante par rapport au dataset. 10h30 : Réseaux de neurones en goulot pour la Reconnaissance du Locuteur – GUILLAUME DAVAL-FREROT Le réseau de neurones en goulot est une forme particulière de réseau de neurones profond. Ce travail présente son principe et son utilisation pour re-traiter les Coefficients MFCC dans un système de reconnaissance du locuteur, pour ensuite étudier les performances du réseau seul et au sein du système. 11h00 : Calcul de distance par un réseau de neurones profond siamois – ANATOLE MOREAU Nous utilisons une architecture siamoise pour calculer une similarité pour une paire d’échantillons. Le DNN projète ces échantillons dans un sous-espace à dimension réduite. On compare les résultats avec les mesures classiques. SPOT — BIBLIOTHÈQUE DE MODEL CHECKING 13h00 : Le support des automates alternants – AMAURY FAUCHILLE Les automates alternants ajoutent un branchement universel au branchement existentiel des automates non-déterministes. Les automates alternants de Büchi sont exponentiellement plus concis que les automates de Büchi non-déterministes. De plus, ils conviennent bien à la traduction de la logique temporelle linéaire car leur taille est proportionnelle à la taille de la formule traduite. Ce travail vise à ajouter le support des automates alternants à Spot. Cela rendra Spot compatible avec les outils produisant et utilisant des automates alternants, et permettra à de futurs algorithmes travaillant sur les automates alternants d’être implémentés. 13h30 : Produit d’automates à parité – LAURENT XU Spot peut construire des automates à parité déterministes à partir d’automates de Büchi non-déterministes. Nous présentons une méthode de produit d’automates à parité basés sur les transitions, qui conserve la parité. Cette méthode est susceptible d’améliorer la déterminisation d’automate existante. OLENA — TRAITEMENT D’IMAGES GÉNÉ- RIQUE ET PERFORMANT 14h00 : Découpage automatique des cartes de Cassini – ANNE-CLAIRE BERTHET Nous cherchons à segmenter d’anciennes cartes de France qui ont été découpées en morceaux puis recollées sur du tissu. Grâce à des filtres morphologiques, on obtient des parties de la frontière au pixel près. Pour finaliser la segmentation, on définit un masque séparant l’image en trois zones qui, à l’aide d’une méthode de classification, permet d’obtenir des résultats proches de la frontière réelle. 14h30 : Discrimination supervisée de caractères sur des images – THIBAULT DEUTSCH La séparation lettre/non-lettre est un sujet important dans le domaine de la reconnaissance de caractères. Afin de générer un modèle basé sur les données d’entraînement, nous analysons les composantes principales (PCA), puis nous appliquons une PLDA. VCSN — BIBLIOTHÈQUE DE MANIPULATION D’AUTOMATES 15h30 : Génération aléatoire d’expression rationnelle – LUCIEN BOILLOD Nous présentons l’implémentation d’un algorithme efficace et générique de génération aléatoire d’expressions rationnelles pondérées. Nous présentons ensuite une manière de générer des chemins aléatoires dans des automates pondérés. 16h00 : Quotients d’automates pondérés et de séries rationnelles – THIBAUD MICHAUD Nous montrons comment le quotient a été implémenté dans Vcsn. Après avoir défini le quotient à gauche et à droite de langages, nous expliquons l’algorithme implémenté dans Vcsn pour calculer le quotient de deux automates. Nous explorons ensuite les conséquences de l’introduction du quotient du côté des expressions. 16h30 : K plus courts chemins dans Vcsn – SÉBASTIEN PIAT Plusieurs algorithmes existent pour obtenir plusieurs plus courts chemins d’un graphe. L’un d’eux, l’algorithme de Yen, cache des transitions dans le graphe entre chaque calcul de plus court chemin. Ce travail présente l’implementation de cet algorithme dans Vcsn ainsi que les différentes techniques d’optimisation envisagées (notamment l’implémentation d’un ensemble creux). 17h00 : Contribution à dyn : : – RAOUL BILLION Pour palier au manque de flexibilité et au temps de compilation assez lourd du C++ Vcsn introduit une couche nommée dyn : : permettant entre autre un système de compilation à la volée. Ce travail présente nos réflexions sur comment gérer et optimiser nos temps de compilation ainsi que des améliorations apportées à la couche dyn : :. -- Daniela Becker Responsable administrative du LRDE

8 years, 9 months

← Newer
1
...
15
16
17
18
19
20
21
...
57
Older →

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Annonce